You don't prove AI visibility work with one metric — you prove it by tracking three things at once: whether AI engines mention the brand when asked, whether the crawlers can actually read the pages you're pointing them to, and whether Search Console impressions are climbing even when clicks lag. Show an owner all three moving together over weeks, not a single screenshot from a good day, and the skepticism has somewhere honest to land.
The report I pulled this morning, on our own site
Before I write anything for a client, I run the same check on Acromatico's own dashboard, because I'm not going to sell an owner a report I wouldn't trust myself. This morning's pull included a line that would make most business owners want their money back: one of our own pages, a comparison guide on AI visibility tool pricing, had picked up 905 impressions in Search Console over the last 28 days — and exactly 1 click.
That's a 0.1% click-through rate on nearly a thousand impressions. If I showed that number to an owner with no context, they'd be right to be angry. It looks like failure. And this is exactly the moment where most agencies either hide the number, bury it three tabs deep in a PDF, or spin it with a sentence like "impressions are a leading indicator" and hope nobody asks a follow-up question.
I don't do that. I put the number on the screen and explain what it actually means, because the explanation is checkable — and checkable is the only kind of proof that survives a second meeting.
Why a smart owner doesn't trust the first number you show them
Every owner who's hired an agency before has seen a "results" report that was really a highlight reel — the one keyword that jumped, the one week traffic spiked, cropped tight enough that you never see the eleven other weeks it didn't move. That's not lying exactly, it's selective framing, and owners develop a nose for it fast. The tell is always the same: a report with no bad numbers in it.
The fix isn't better spin. It's showing the number that looks bad and having a mechanical, checkable reason for it sitting right next to it. An owner doesn't need every metric to be good. They need to trust that when a number is bad, you'll say so before they find it themselves.
What 905 impressions and 1 click actually means
A page that ranks well enough to be shown 905 times but earns one click is not an invisible page — it's the opposite problem. It's being surfaced. Google is putting it in front of searchers on relevant queries at a real volume. The failure is narrower and more fixable than "the strategy isn't working": either the title and meta description aren't earning the click against what's next to them on the results page, or the query intent is being satisfied by an AI Overview or a snippet before the searcher ever needs to click through.
That's a title/meta rewrite and a positioning check, not a rebuild. It's also exactly the kind of signal that only shows up if you're actually reading Search Console query-by-query instead of reporting one blended "traffic" number, which is the habit we cover in depth here when AI Overviews are eating clicks across a whole site.
The three-number report that survives a skeptical owner
We don't lead with one dashboard screenshot. We show three things side by side, because each one answers a different objection an owner is silently holding.
| What we show | The objection it answers | Where it comes from |
|---|---|---|
| Search Console impressions, by page, trended over 8+ weeks | "Is anything actually happening, or is this just talk?" | Google Search Console Performance report |
| A curl check confirming AI crawlers can read the page that's supposed to rank | "How do I know the AI engines can even see what you built?" | Server response headers for GPTBot, ClaudeBot, PerplexityBot, Google-Extended — the same 20-minute check in this guide |
| A direct prompt test — asking ChatGPT and Perplexity the target question, screenshotted | "Does any of this actually change what a customer sees when they ask an AI?" | Manual query against the live engines, on a fixed cadence |
None of these three, alone, is proof. A crawler being able to read a page doesn't mean it ranks. A prompt test being favorable one week doesn't mean it holds the next. But an owner who sees impressions climbing, crawlers confirmed unblocked, and a prompt test that's trending toward a mention over 60 days has three independent lines of evidence pointing the same direction — which is a much harder thing to dismiss than a single chart.
The line I actually say in the room
When an owner points at a bad number, I don't defend it. I say something close to this: "That number is real, and here's exactly what it's telling us — not what it's hiding." Then I show the mechanism, not a promise. For the 905/1 page, that meant pulling up the actual query list underneath it, showing which searches were triggering the impressions, and pointing at the title tag that was competing against three other results with nearly identical wording.
The owner's actual question is rarely "is the number good." It's "can I trust the person telling me about the number." Those are different questions, and only one of them gets answered by hiding the bad ones.
Setting the timeline before the work starts, not after
A lot of the skepticism we walk into isn't about the work — it's about a previous agency setting no timeline at all, so every month became a moving target. We set it explicitly, before a contract is signed:
- Weeks 1–2: baseline pull — current impressions, current crawler access, current AI mention rate on the owner's top 5–10 buyer questions. This is the number everything else gets measured against.
- Weeks 3–6: Search Console impressions are the first thing to move, as Google recrawls and re-ranks updated and new pages. This is genuinely the earliest honest signal — anyone promising ranking movement in week one is either lucky or not being straight with you.
- Day 60: the first real review. AI citation movement lags Search Console because retrieval-augmented engines and cached training data refresh on their own schedule, not the client's. We tell owners explicitly not to judge citation progress before this point.
- Day 90+: click-through and conversion movement, assuming the title/meta and positioning work from the earlier weeks has had time to compound.
Writing this down before the work starts does something a report can't do after the fact: it turns "why isn't this working yet" into "we're on week 4 of a 60-day cycle we agreed to," which is a completely different conversation.
What we do the moment we see a number like this
The 905/1 page didn't sit on a to-do list. It became today's queued fix: a title and meta rewrite targeting the actual query intent behind the impressions, checked against what's currently ranking above and beside it. That's the discipline that matters more than any single report — a bad number gets acted on the same week it's found, not explained away and revisited next quarter. It's the same standard behind the before/after teardown we published on rewriting an underperforming page for citation instead of pure ranking.
Questions people ask
You track three separate things, not one blended number: whether AI engines mention the brand at all when asked relevant questions, whether ranking pages are actually being read by AI crawlers (verified in server logs, not assumed), and whether Search Console shows impressions climbing even if clicks lag behind them. A skeptical owner needs to see all three move together over weeks, not a single vanity metric from one week.
It usually means the page is being surfaced — in a Google AI Overview, a broad informational search, or a low-intent query — but the title and meta description aren't earning the click, or the page is answering a question the searcher already got answered on the results page itself. High impressions with a near-zero CTR is a real, checkable signal, not a failure of the whole strategy; it tells you exactly which page to rewrite next.
Search Console impressions on new or improved pages typically move within two to six weeks as engines recrawl and re-rank. AI engine citations lag further behind because retrieval-augmented systems and cached training data refresh on their own schedules, not yours. We tell owners to expect visible movement in Search Console first, citation movement second, and set the first real review at 60 days rather than 30.
Want the same three-number report run on your own site?
A free audit shows exactly where you're getting impressions with no clicks, whether AI crawlers can even read your pages, and what AI engines say about you today.
Get a free AI Visibility Audit →